深度学习每周学习总结J1（ResNet-50算法实战与解析

🍨 本文为🔗365天深度学习训练营中的学习记录博客
🍖 原作者：K同学啊 | 接辅导、项目定制

- - 0. 总结
  - 1. 设置GPU
  - 2. 导入数据及处理部分
  - 3. 划分数据集
  - 4. 模型构建部分
  - 5. 设置超参数：定义损失函数，学习率，以及根据学习率定义优化器等
  - 6. 训练函数
  - 7. 测试函数
  - 7. 正式训练
  - 9. 结果可视化
  - 10. 模型的保存
  - 11. 使用训练好的模型进行预测

0. 总结

数据导入及处理部分：本次数据导入没有使用torchvision自带的数据集，需要将原始数据进行处理包括数据导入，查看数据分类情况，定义transforms，进行数据类型转换等操作。

划分数据集：划定训练集测试集后，再使用torch.utils.data中的DataLoader()分别加载上一步处理好的训练及测试数据，查看批处理维度.

模型构建部分：resnet-50

设置超参数：在这之前需要定义损失函数，学习率（动态学习率），以及根据学习率定义优化器（例如SGD随机梯度下降），用来在训练中更新参数，最小化损失函数。

定义训练函数：函数的传入的参数有四个，分别是设置好的DataLoader(),定义好的模型，损失函数，优化器。函数内部初始化损失准确率为0，接着开始循环，使用DataLoader()获取一个批次的数据，对这个批次的数据带入模型得到预测值，然后使用损失函数计算得到损失值。接下来就是进行反向传播以及使用优化器优化参数，梯度清零放在反向传播之前或者是使用优化器优化之后都是可以的，一般是默认放在反向传播之前。

定义测试函数：函数传入的参数相比训练函数少了优化器，只需传入设置好的DataLoader(),定义好的模型，损失函数。此外除了处理批次数据时无需再设置梯度清零、返向传播以及优化器优化参数，其余部分均和训练函数保持一致。

训练过程：定义训练次数，有几次就使用整个数据集进行几次训练，初始化四个空list分别存储每次训练及测试的准确率及损失。使用model.train()开启训练模式，调用训练函数得到准确率及损失。使用model.eval()将模型设置为评估模式，调用测试函数得到准确率及损失。接着就是将得到的训练及测试的准确率及损失存储到相应list中并合并打印出来，得到每一次整体训练后的准确率及损失。

结果可视化

模型的保存，调取及使用。在PyTorch中，通常使用 torch.save(model.state_dict(), ‘model.pth’) 保存模型的参数，使用 model.load_state_dict(torch.load(‘model.pth’)) 加载参数。

需要改进优化的地方：在保证整体流程没有问题的情况下，继续细化细节研究，比如一些函数的原理及作用，如何提升训练集准确率等问题。

import torch
import torch.nn as nn
import torchvision
from torchvision import datasets,transforms
from torch.utils.data import DataLoader
import torchvision.models as models
import torch.nn.functional as Fimport os,PIL,pathlib
import matplotlib.pyplot as plt
import warningswarnings.filterwarnings('ignore') # 忽略警告信息plt.rcParams['font.sans-serif'] = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False   # 用来正常显示负号
plt.rcParams['figure.dpi'] = 100 # 分辨率

1. 设置GPU

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device

device(type='cuda')

2. 导入数据及处理部分

# 获取数据分布情况
path_dir = './data/bird_photos/'
path_dir = pathlib.Path(path_dir)paths = list(path_dir.glob('*'))
# classNames = [str(path).split("\\")[-1] for path in paths] # ['Bananaquit', 'Black Skimmer', 'Black Throated Bushtiti', 'Cockatoo']
classNames = [path.parts[-1] for path in paths]
classNames

['Bananaquit', 'Black Skimmer', 'Black Throated Bushtiti', 'Cockatoo']

# 定义transforms 并处理数据
train_transforms = transforms.Compose([transforms.Resize([224,224]),      # 将输入图片resize成统一尺寸transforms.RandomHorizontalFlip(), # 随机水平翻转transforms.ToTensor(),             # 将PIL Image 或 numpy.ndarray 装换为tensor,并归一化到[0,1]之间transforms.Normalize(              # 标准化处理 --> 转换为标准正太分布（高斯分布），使模型更容易收敛mean = [0.485,0.456,0.406],    # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。std = [0.229,0.224,0.225])
])
test_transforms = transforms.Compose([transforms.Resize([224,224]),transforms.ToTensor(),transforms.Normalize(mean = [0.485,0.456,0.406],std = [0.229,0.224,0.225])
])
total_data = datasets.ImageFolder('./data/bird_photos/',transform = train_transforms)
total_data

Dataset ImageFolderNumber of datapoints: 565Root location: ./data/bird_photos/StandardTransform
Transform: Compose(Resize(size=[224, 224], interpolation=bilinear, max_size=None, antialias=True)RandomHorizontalFlip(p=0.5)ToTensor()Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]))

total_data.class_to_idx

{'Bananaquit': 0,'Black Skimmer': 1,'Black Throated Bushtiti': 2,'Cockatoo': 3}

3. 划分数据集

# 划分数据集
train_size = int(len(total_data) * 0.8)
test_size = len(total_data) - train_sizetrain_dataset,test_dataset = torch.utils.data.random_split(total_data,[train_size,test_size])
train_dataset,test_dataset

(<torch.utils.data.dataset.Subset at 0x23b545994b0>,<torch.utils.data.dataset.Subset at 0x23b54599300>)

# 定义DataLoader用于数据集的加载batch_size = 32train_dl = torch.utils.data.DataLoader(train_dataset,batch_size = batch_size,shuffle = True,num_workers = 1
)
test_dl = torch.utils.data.DataLoader(test_dataset,batch_size = batch_size,shuffle = True,num_workers = 1
)

# 观察数据维度
for X,y in test_dl:print("Shape of X [N,C,H,W]: ",X.shape)print("Shape of y: ", y.shape,y.dtype)break

Shape of X [N,C,H,W]:  torch.Size([32, 3, 224, 224])
Shape of y:  torch.Size([32]) torch.int64

4. 模型构建部分

import torch
import torch.nn as nn
import torch.nn.functional as Fclass ConvBlock(nn.Module):def __init__(self, in_channels, filters, kernel_size, strides=1):super(ConvBlock, self).__init__()filters1, filters2, filters3 = filtersself.conv1 = nn.Conv2d(in_channels, filters1, kernel_size=1, stride=strides)self.bn1 = nn.BatchNorm2d(filters1)self.conv2 = nn.Conv2d(filters1, filters2, kernel_size=kernel_size, padding=1)self.bn2 = nn.BatchNorm2d(filters2)self.conv3 = nn.Conv2d(filters2, filters3, kernel_size=1)self.bn3 = nn.BatchNorm2d(filters3)self.shortcut = nn.Sequential()if in_channels != filters3 or strides != 1:self.shortcut = nn.Sequential(nn.Conv2d(in_channels, filters3, kernel_size=1, stride=strides),nn.BatchNorm2d(filters3))def forward(self, x):shortcut = self.shortcut(x)x = F.relu(self.bn1(self.conv1(x)))x = F.relu(self.bn2(self.conv2(x)))x = self.bn3(self.conv3(x))x += shortcutx = F.relu(x)return xclass ResNet50(nn.Module):def __init__(self, num_classes=1000):super(ResNet50, self).__init__()self.pad = nn.ZeroPad2d(3)self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3)self.bn1 = nn.BatchNorm2d(64)self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)# Define block layers with appropriate strides and filter sizesself.layer1 = self._build_layer(64, [64, 64, 256], blocks=3, strides=1)self.layer2 = self._build_layer(256, [128, 128, 512], blocks=4, strides=2)self.layer3 = self._build_layer(512, [256, 256, 1024], blocks=6, strides=2)self.layer4 = self._build_layer(1024, [512, 512, 2048], blocks=3, strides=2)self.avgpool = nn.AdaptiveAvgPool2d((1, 1))self.fc = nn.Linear(2048, num_classes)def _build_layer(self, in_channels, filters, blocks, strides=1):layers = []# Add the first block with potential downsampling (by strides)layers.append(ConvBlock(in_channels, filters, kernel_size=3, strides=strides))in_channels = filters[-1]# Add the additional blocks in this layer without downsamplingfor _ in range(1, blocks):layers.append(ConvBlock(in_channels, filters, kernel_size=3, strides=1))return nn.Sequential(*layers)def forward(self, x):x = self.pad(x)x = F.relu(self.bn1(self.conv1(x)))x = self.maxpool(x)x = self.layer1(x)x = self.layer2(x)x = self.layer3(x)x = self.layer4(x)x = self.avgpool(x)x = torch.flatten(x, 1)x = self.fc(x)return x# Ensure the model uses the correct device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = ResNet50()
model.to(device)

ResNet50((pad): ZeroPad2d((3, 3, 3, 3))(conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3))(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)(layer1): Sequential((0): ConvBlock((conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential((0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): ConvBlock((conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential())(2): ConvBlock((conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential()))(layer2): Sequential((0): ConvBlock((conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(2, 2))(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential((0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2))(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): ConvBlock((conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential())(2): ConvBlock((conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential())(3): ConvBlock((conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential()))(layer3): Sequential((0): ConvBlock((conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(2, 2))(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential((0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2))(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): ConvBlock((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential())(2): ConvBlock((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential())(3): ConvBlock((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential())(4): ConvBlock((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential())(5): ConvBlock((conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential()))(layer4): Sequential((0): ConvBlock((conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(2, 2))(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential((0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2))(1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)))(1): ConvBlock((conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential())(2): ConvBlock((conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1))(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1))(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)(shortcut): Sequential()))(avgpool): AdaptiveAvgPool2d(output_size=(1, 1))(fc): Linear(in_features=2048, out_features=1000, bias=True)
)

5. 设置超参数：定义损失函数，学习率，以及根据学习率定义优化器等

# loss_fn = nn.CrossEntropyLoss() # 创建损失函数# learn_rate = 1e-3 # 初始学习率
# def adjust_learning_rate(optimizer,epoch,start_lr):
#     # 每两个epoch 衰减到原来的0.98
#     lr = start_lr * (0.92 ** (epoch//2))
#     for param_group in optimizer.param_groups:
#         param_group['lr'] = lr# optimizer = torch.optim.Adam(model.parameters(),lr=learn_rate)

# 调用官方接口示例
loss_fn = nn.CrossEntropyLoss()learn_rate = 1e-4
lambda1 = lambda epoch:(0.92**(epoch//2))optimizer = torch.optim.Adam(model.parameters(),lr = learn_rate)
scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer,lr_lambda=lambda1) # 选定调整方法

6. 训练函数

# 训练函数
def train(dataloader,model,loss_fn,optimizer):size = len(dataloader.dataset) # 训练集大小num_batches = len(dataloader) # 批次数目train_loss,train_acc = 0,0for X,y in dataloader:X,y = X.to(device),y.to(device)# 计算预测误差pred = model(X)loss = loss_fn(pred,y)# 反向传播optimizer.zero_grad()loss.backward()optimizer.step()# 记录acc与losstrain_acc += (pred.argmax(1)==y).type(torch.float).sum().item()train_loss += loss.item()train_acc /= sizetrain_loss /= num_batchesreturn train_acc,train_loss

7. 测试函数

# 测试函数
def test(dataloader,model,loss_fn):size = len(dataloader.dataset)num_batches = len(dataloader)test_acc,test_loss = 0,0with torch.no_grad():for X,y in dataloader:X,y = X.to(device),y.to(device)# 计算losspred = model(X)loss = loss_fn(pred,y)test_acc += (pred.argmax(1)==y).type(torch.float).sum().item()test_loss += loss.item()test_acc /= sizetest_loss /= num_batchesreturn test_acc,test_loss

7. 正式训练

import copyepochs = 40train_acc = []
train_loss = []
test_acc = []
test_loss = []best_acc = 0.0for epoch in range(epochs):# 更新学习率——使用自定义学习率时使用# adjust_learning_rate(optimizer,epoch,learn_rate)model.train()epoch_train_acc,epoch_train_loss = train(train_dl,model,loss_fn,optimizer)scheduler.step() # 更新学习率——调用官方动态学习率时使用model.eval()epoch_test_acc,epoch_test_loss = test(test_dl,model,loss_fn)# 保存最佳模型到 best_modelif epoch_test_acc > best_acc:best_acc = epoch_test_accbest_model = copy.deepcopy(model)train_acc.append(epoch_train_acc)train_loss.append(epoch_train_loss)test_acc.append(epoch_test_acc)test_loss.append(epoch_test_loss)# 获取当前学习率lr = optimizer.state_dict()['param_groups'][0]['lr']template = ('Epoch:{:2d},Train_acc:{:.1f}%,Train_loss:{:.3f},Test_acc:{:.1f}%,Test_loss:{:.3f},Lr:{:.2E}')print(template.format(epoch+1,epoch_train_acc*100,epoch_train_loss,epoch_test_acc*100,epoch_test_loss,lr))print('Done')

Epoch: 1,Train_acc:35.4%,Train_loss:3.431,Test_acc:23.9%,Test_loss:3.033,Lr:1.00E-04
Epoch: 2,Train_acc:69.2%,Train_loss:0.899,Test_acc:26.5%,Test_loss:3.017,Lr:9.20E-05
Epoch: 3,Train_acc:74.1%,Train_loss:0.652,Test_acc:59.3%,Test_loss:1.278,Lr:9.20E-05
Epoch: 4,Train_acc:78.3%,Train_loss:0.587,Test_acc:67.3%,Test_loss:1.353,Lr:8.46E-05
Epoch: 5,Train_acc:82.5%,Train_loss:0.521,Test_acc:75.2%,Test_loss:0.829,Lr:8.46E-05
Epoch: 6,Train_acc:86.3%,Train_loss:0.433,Test_acc:66.4%,Test_loss:1.308,Lr:7.79E-05
Epoch: 7,Train_acc:90.3%,Train_loss:0.340,Test_acc:67.3%,Test_loss:1.600,Lr:7.79E-05
Epoch: 8,Train_acc:89.2%,Train_loss:0.374,Test_acc:71.7%,Test_loss:1.014,Lr:7.16E-05
Epoch: 9,Train_acc:88.7%,Train_loss:0.305,Test_acc:77.0%,Test_loss:0.841,Lr:7.16E-05
Epoch:10,Train_acc:90.7%,Train_loss:0.309,Test_acc:79.6%,Test_loss:1.094,Lr:6.59E-05
Epoch:11,Train_acc:91.8%,Train_loss:0.318,Test_acc:72.6%,Test_loss:0.976,Lr:6.59E-05
Epoch:12,Train_acc:93.6%,Train_loss:0.243,Test_acc:73.5%,Test_loss:1.209,Lr:6.06E-05
Epoch:13,Train_acc:94.7%,Train_loss:0.159,Test_acc:71.7%,Test_loss:0.947,Lr:6.06E-05
Epoch:14,Train_acc:98.0%,Train_loss:0.076,Test_acc:80.5%,Test_loss:0.707,Lr:5.58E-05
Epoch:15,Train_acc:97.8%,Train_loss:0.083,Test_acc:79.6%,Test_loss:0.923,Lr:5.58E-05
Epoch:16,Train_acc:98.2%,Train_loss:0.059,Test_acc:82.3%,Test_loss:0.650,Lr:5.13E-05
Epoch:17,Train_acc:98.5%,Train_loss:0.072,Test_acc:76.1%,Test_loss:0.828,Lr:5.13E-05
Epoch:18,Train_acc:98.0%,Train_loss:0.175,Test_acc:78.8%,Test_loss:0.834,Lr:4.72E-05
Epoch:19,Train_acc:94.2%,Train_loss:0.173,Test_acc:61.9%,Test_loss:2.606,Lr:4.72E-05
Epoch:20,Train_acc:96.2%,Train_loss:0.123,Test_acc:77.9%,Test_loss:0.959,Lr:4.34E-05
Epoch:21,Train_acc:96.5%,Train_loss:0.166,Test_acc:76.1%,Test_loss:1.266,Lr:4.34E-05
Epoch:22,Train_acc:96.9%,Train_loss:0.196,Test_acc:85.0%,Test_loss:0.698,Lr:4.00E-05
Epoch:23,Train_acc:98.7%,Train_loss:0.082,Test_acc:82.3%,Test_loss:0.626,Lr:4.00E-05
Epoch:24,Train_acc:96.0%,Train_loss:0.136,Test_acc:81.4%,Test_loss:0.805,Lr:3.68E-05
Epoch:25,Train_acc:98.5%,Train_loss:0.101,Test_acc:83.2%,Test_loss:0.576,Lr:3.68E-05
Epoch:26,Train_acc:98.5%,Train_loss:0.062,Test_acc:80.5%,Test_loss:0.597,Lr:3.38E-05
Epoch:27,Train_acc:99.6%,Train_loss:0.039,Test_acc:83.2%,Test_loss:0.574,Lr:3.38E-05
Epoch:28,Train_acc:99.3%,Train_loss:0.080,Test_acc:85.0%,Test_loss:0.758,Lr:3.11E-05
Epoch:29,Train_acc:99.6%,Train_loss:0.059,Test_acc:84.1%,Test_loss:0.608,Lr:3.11E-05
Epoch:30,Train_acc:98.9%,Train_loss:0.054,Test_acc:82.3%,Test_loss:0.753,Lr:2.86E-05
Epoch:31,Train_acc:98.9%,Train_loss:0.035,Test_acc:83.2%,Test_loss:0.617,Lr:2.86E-05
Epoch:32,Train_acc:98.7%,Train_loss:0.046,Test_acc:78.8%,Test_loss:0.847,Lr:2.63E-05
Epoch:33,Train_acc:98.9%,Train_loss:0.028,Test_acc:82.3%,Test_loss:0.746,Lr:2.63E-05
Epoch:34,Train_acc:99.6%,Train_loss:0.032,Test_acc:79.6%,Test_loss:0.629,Lr:2.42E-05
Epoch:35,Train_acc:99.3%,Train_loss:0.027,Test_acc:82.3%,Test_loss:0.597,Lr:2.42E-05
Epoch:36,Train_acc:99.6%,Train_loss:0.029,Test_acc:87.6%,Test_loss:0.488,Lr:2.23E-05
Epoch:37,Train_acc:99.6%,Train_loss:0.026,Test_acc:87.6%,Test_loss:0.552,Lr:2.23E-05
Epoch:38,Train_acc:99.1%,Train_loss:0.029,Test_acc:79.6%,Test_loss:0.572,Lr:2.05E-05
Epoch:39,Train_acc:99.8%,Train_loss:0.107,Test_acc:84.1%,Test_loss:0.704,Lr:2.05E-05
Epoch:40,Train_acc:99.1%,Train_loss:0.094,Test_acc:76.1%,Test_loss:0.772,Lr:1.89E-05
Done

9. 结果可视化

epochs_range = range(epochs)plt.figure(figsize = (12,3))plt.subplot(1,2,1)
plt.plot(epochs_range,train_acc,label = 'Training Accuracy')
plt.plot(epochs_range,test_acc,label = 'Test Accuracy')
plt.legend(loc = 'lower right')
plt.title('Training and Validation Accuracy')plt.subplot(1,2,2)
plt.plot(epochs_range,train_loss,label = 'Test Accuracy')
plt.plot(epochs_range,test_loss,label = 'Test Loss')
plt.legend(loc = 'lower right')
plt.title('Training and validation Loss')
plt.show()

在这里插入图片描述

10. 模型的保存

# 自定义模型保存
# 状态字典保存
torch.save(model.state_dict(),'./模型参数/J1_resnet50_model_state_dict.pth') # 仅保存状态字典# 加载状态字典到模型
best_model = ResNet50().to(device) # 定义官方vgg16模型用来加载参数best_model.load_state_dict(torch.load('./模型参数/J1_resnet50_model_state_dict.pth')) # 加载状态字典到模型

<All keys matched successfully>

11. 使用训练好的模型进行预测

# 指定路径图片预测
from PIL import Image
import torchvision.transforms as transformsclasses = list(total_data.class_to_idx) # classes = list(total_data.class_to_idx)def predict_one_image(image_path,model,transform,classes):test_img = Image.open(image_path).convert('RGB')# plt.imshow(test_img) # 展示待预测的图片test_img = transform(test_img)img = test_img.to(device).unsqueeze(0)model.eval()output = model(img)print(output) # 观察模型预测结果的输出数据_,pred = torch.max(output,1)pred_class = classes[pred]print(f'预测结果是:{pred_class}')

# 预测训练集中的某张照片
predict_one_image(image_path='./data/bird_photos/Bananaquit/007.jpg',model = model,transform = test_transforms,classes = classes)

tensor([[13.8683,  4.9143,  4.8488,  0.3971, -4.3269, -6.6820, -5.4773, -4.8104,-6.3061, -6.0159, -4.1786, -5.2478, -4.7984, -4.5901, -5.0879, -4.8318,-5.1802, -4.8213, -4.1029, -4.8857, -5.7887, -4.4304, -5.2544, -4.4558,-3.9174, -4.3528, -5.6769, -5.0493, -4.5999, -6.1482, -4.0200, -3.2741,-4.6080, -6.3622, -3.1413, -4.3438, -5.2021, -4.2925, -5.5243, -4.7640,-4.9537, -5.5491, -5.6634, -3.9548, -4.3976, -5.2634, -5.1851, -4.0045,-6.0956, -4.2346, -4.1718, -4.4907, -4.1201, -4.3656, -4.0707, -5.3282,-5.3788, -4.4176, -4.8130, -5.5919, -5.5178, -5.1131, -4.6648, -4.9041,-5.3978, -5.4855, -5.0189, -4.8627, -5.7097, -4.2608, -5.3921, -4.4241,-5.8152, -5.2379, -5.2549, -5.9455, -4.4937, -5.2863, -5.2548, -5.5048,-4.9770, -5.0247, -4.8115, -5.1978, -5.0874, -4.8657, -4.7004, -6.1652,-7.1067, -5.3773, -5.5511, -4.6870, -5.7534, -4.9669, -4.2131, -5.3542,-4.9979, -5.4464, -5.2461, -4.8821, -5.5314, -5.8627, -4.4703, -4.0505,-4.3990, -6.0654, -4.9767, -4.8933, -5.0749, -5.0969, -4.7121, -4.7877,-5.9138, -5.0609, -5.6325, -4.4758, -5.3832, -5.3308, -5.1663, -5.3744,-4.6568, -5.5547, -4.7015, -4.5332, -4.7048, -4.0575, -4.9354, -5.9719,-4.3700, -4.3375, -4.3835, -5.3360, -4.6949, -4.7043, -3.0128, -5.4335,-4.1490, -5.5024, -5.1213, -4.5514, -4.3888, -5.1336, -4.9035, -2.9031,-4.2921, -5.1654, -4.5989, -5.7990, -5.1267, -4.8723, -5.2425, -5.2648,-4.7031, -4.4419, -4.9500, -5.4625, -4.4614, -6.0281, -4.5982, -5.2243,-5.2080, -4.5609, -4.2382, -5.6643, -5.3306, -4.9993, -5.4132, -5.1988,-4.8110, -4.3645, -4.7883, -4.4945, -4.1371, -6.1615, -4.3348, -6.0694,-5.0832, -5.7544, -7.4370, -5.0871, -4.8265, -4.9912, -5.4170, -4.7337,-4.8869, -4.2867, -6.0734, -5.3458, -4.6136, -4.6427, -5.0670, -5.2674,-5.3138, -6.1891, -4.5351, -5.4272, -4.0822, -3.8868, -5.3524, -4.5716,-5.2943, -5.4524, -5.8091, -4.6113, -6.2336, -3.9732, -5.5085, -4.9551,-5.8190, -4.7829, -6.0425, -4.4670, -5.0961, -4.8832, -4.9884, -5.3502,-4.7928, -4.6363, -5.2002, -5.2305, -6.3564, -4.8475, -5.7222, -4.9972,-4.8803, -4.9031, -4.7486, -5.1633, -4.4830, -5.8250, -4.9781, -6.4707,-5.1635, -6.2186, -4.6911, -6.1466, -4.3560, -4.0647, -5.6964, -4.6971,-4.5650, -5.2661, -4.4342, -6.5263, -5.0083, -4.1184, -4.4144, -4.3142,-3.9152, -4.6833, -3.9448, -4.7297, -4.7736, -4.5994, -5.2097, -3.7666,-4.4968, -4.7199, -3.7594, -4.8945, -4.8153, -4.3126, -5.4928, -4.2669,-4.4810, -5.1798, -6.5569, -6.4180, -6.2858, -5.0941, -4.5702, -5.1153,-5.6136, -5.2451, -4.9205, -5.2890, -4.1699, -4.1866, -4.2302, -4.9986,-4.8075, -5.6134, -3.6785, -5.4610, -5.5478, -6.9582, -4.6375, -5.1448,-4.6614, -5.0536, -4.4895, -5.3882, -4.4433, -5.9130, -5.5548, -5.1081,-5.9129, -5.2355, -5.0405, -4.8093, -4.5147, -6.5717, -4.7978, -6.3876,-4.6591, -6.3384, -5.6406, -5.2991, -3.7750, -4.6432, -5.0256, -5.9266,-4.7254, -5.3511, -4.1279, -5.8108, -5.6086, -4.8933, -4.4658, -4.2604,-4.2203, -6.0904, -4.4941, -4.3299, -4.4338, -4.5120, -5.5362, -5.6916,-4.3631, -5.7946, -4.5226, -5.0334, -5.3108, -4.6992, -5.8546, -5.3448,-4.7411, -4.6165, -5.0561, -3.7938, -4.5349, -6.3409, -5.6717, -5.2238,-5.4950, -5.6322, -5.9336, -4.7374, -5.2193, -4.7341, -5.5894, -4.9538,-4.9265, -4.3776, -5.9501, -4.6434, -4.7603, -3.4144, -4.9672, -4.5454,-4.5984, -4.3427, -5.4971, -4.6948, -4.5390, -4.1352, -5.3147, -5.4603,-5.7149, -4.4007, -5.0655, -5.0953, -3.7071, -5.5659, -4.5470, -4.9357,-4.5709, -4.9261, -4.6197, -4.9192, -5.8589, -5.4089, -4.1649, -5.2944,-5.5576, -5.7486, -4.8553, -4.2282, -4.0010, -6.0522, -5.6085, -5.3450,-5.6549, -4.8746, -4.5044, -5.1996, -5.0918, -4.1159, -4.2053, -4.7821,-5.6509, -6.5732, -4.7520, -5.1714, -5.1409, -5.1455, -6.2418, -5.0210,-5.5984, -3.9693, -5.4372, -5.1220, -5.2193, -4.5396, -4.4132, -5.4393,-4.9878, -5.6195, -4.7220, -5.4855, -4.8570, -4.3293, -5.4886, -5.1216,-6.0905, -5.6525, -4.9218, -5.5533, -4.9065, -5.2826, -4.3511, -4.3716,-4.7577, -5.2228, -4.3568, -5.2613, -7.0981, -4.7413, -4.1612, -5.1176,-5.5265, -4.8816, -5.5066, -3.1811, -5.5949, -5.1209, -5.3330, -4.5807,-5.0738, -6.2544, -5.7920, -4.8863, -5.4842, -4.1021, -4.2780, -5.6989,-4.9062, -4.9241, -4.6026, -3.8766, -4.2304, -5.3913, -3.9213, -4.8042,-3.8525, -4.7163, -5.0682, -5.0261, -5.7655, -6.6659, -5.0076, -4.4935,-4.9612, -5.2125, -5.8215, -5.3559, -4.8470, -4.5751, -5.2267, -5.7606,-5.1162, -5.7072, -4.0790, -5.3081, -4.5741, -5.2311, -5.0682, -3.9425,-4.0050, -5.5297, -4.4969, -4.0187, -5.7901, -4.9107, -5.2692, -4.6438,-5.0574, -5.8231, -5.3713, -5.0068, -4.3541, -4.4603, -4.4939, -4.8495,-5.5300, -4.9960, -3.5204, -4.3364, -6.1883, -4.1293, -5.0230, -4.7546,-5.8216, -4.1616, -5.9442, -5.3490, -3.8164, -5.5285, -4.4659, -3.8073,-4.0800, -4.3338, -4.8140, -5.4170, -5.5654, -3.5721, -5.4222, -5.4133,-5.3061, -5.6210, -5.0620, -4.9480, -5.4236, -5.0383, -4.5365, -6.5556,-4.9221, -5.6572, -5.3458, -4.0751, -3.5474, -5.1584, -5.4831, -5.0386,-5.1952, -5.7538, -5.2630, -4.6680, -5.3158, -4.4897, -5.2653, -5.6185,-5.2754, -4.9732, -3.6142, -3.3807, -5.9308, -3.3762, -4.5909, -4.8592,-5.4307, -5.8186, -3.1528, -5.8814, -4.3251, -5.4645, -5.7290, -5.5545,-6.9818, -4.6826, -5.2847, -5.2316, -5.3749, -5.5513, -5.0912, -6.0087,-5.0094, -5.2834, -5.9248, -4.2985, -5.6460, -4.0210, -5.0835, -5.0239,-4.2798, -4.0711, -3.0299, -4.4249, -3.8580, -5.4938, -6.1500, -6.2026,-5.5107, -4.6297, -5.2490, -4.1693, -5.5621, -5.3462, -4.4608, -5.4131,-4.8175, -4.7294, -5.3795, -4.7759, -4.9308, -4.6114, -4.8512, -5.6910,-4.5780, -6.0506, -4.5082, -5.0415, -5.1643, -4.3208, -5.7007, -5.4402,-5.0097, -3.7647, -5.9034, -4.9685, -4.2927, -3.3002, -6.0967, -4.7784,-5.0862, -4.9967, -5.5023, -4.9356, -5.8725, -4.7621, -4.8660, -5.6542,-4.1783, -5.8236, -4.4004, -5.2087, -6.2577, -5.9747, -3.9065, -4.9602,-4.9865, -6.0940, -5.1763, -4.1279, -4.7915, -4.5835, -4.4909, -5.0032,-4.4306, -5.1279, -5.9237, -5.6016, -4.1580, -5.8993, -5.5826, -5.1257,-5.2317, -5.4553, -5.0349, -5.3762, -4.8476, -5.7886, -4.9892, -5.3761,-6.2784, -4.3499, -5.6478, -5.4893, -3.0633, -4.5597, -5.5055, -5.4887,-5.8181, -5.6683, -6.3172, -4.5253, -5.2920, -4.5857, -5.5727, -4.5398,-4.7678, -5.5095, -3.8584, -5.2461, -5.6084, -4.6665, -5.4107, -4.8152,-5.4975, -5.0608, -4.3795, -6.9169, -5.0226, -5.4908, -5.5358, -5.0029,-5.5313, -4.9354, -4.8460, -4.9901, -4.7630, -4.8377, -5.0652, -5.9628,-5.9703, -5.4385, -4.6380, -5.6674, -5.3192, -4.4605, -3.6204, -5.3540,-7.2942, -4.0140, -5.5212, -4.0053, -3.5999, -5.2319, -5.9446, -5.6247,-4.3006, -5.1588, -4.5104, -3.9837, -5.1592, -4.2998, -6.2257, -5.3139,-4.5336, -5.6970, -4.5432, -5.0471, -5.1169, -5.7729, -5.3915, -3.4912,-4.6369, -5.2365, -6.0601, -5.3651, -5.0499, -5.1211, -4.7336, -6.7476,-5.6456, -5.0389, -4.7289, -5.5121, -2.9521, -4.6852, -5.8838, -5.6147,-4.9222, -4.7334, -5.0473, -5.3684, -4.3785, -4.4092, -5.7249, -5.1093,-4.9247, -4.0973, -5.2757, -5.0103, -5.3298, -5.2246, -5.2376, -5.0409,-5.3682, -4.4103, -5.2739, -5.7753, -5.3121, -5.3229, -4.7973, -5.3419,-4.0869, -5.5489, -6.0253, -4.6384, -4.7001, -5.3798, -5.4072, -4.6274,-5.2741, -6.1363, -4.6067, -5.3099, -5.0059, -5.1699, -4.9767, -5.2467,-5.0942, -4.4623, -5.6662, -4.6116, -5.3887, -4.7994, -4.1454, -2.2531,-5.6628, -5.2896, -4.7683, -6.1704, -5.5401, -7.0299, -5.1669, -5.5821,-4.7684, -5.2925, -4.6315, -4.8803, -5.1687, -5.2941, -4.9299, -4.8190,-5.4344, -4.7551, -4.7198, -5.1019, -5.7492, -6.1550, -5.6996, -4.4132,-6.2275, -5.4198, -4.8401, -5.1504, -5.1971, -5.7264, -4.6849, -5.0585,-3.8531, -3.6110, -4.3900, -4.0146, -5.6333, -5.7491, -4.2836, -4.9333,-5.2104, -4.4378, -4.6586, -6.6317, -5.3482, -4.9060, -4.0507, -4.3850,-4.3232, -4.4060, -4.6808, -6.1481, -5.0341, -5.0642, -4.2104, -4.4937,-5.8942, -5.7710, -5.0782, -4.6604, -6.3730, -7.0023, -5.4708, -6.7699,-4.6094, -5.3579, -6.3511, -5.2445, -3.5320, -4.9982, -5.1565, -5.1640,-5.4871, -3.6589, -3.9923, -5.0592, -4.5019, -5.9001, -4.1611, -4.6703,-5.1029, -2.9716, -5.3578, -5.1678, -4.0690, -6.2209, -4.3308, -4.6219,-5.7228, -3.6582, -4.6460, -5.5739, -4.9900, -5.2182, -5.1610, -4.2914,-3.4296, -4.7841, -5.4836, -4.0942, -5.3494, -4.3294, -6.1656, -5.4277,-5.9730, -3.6229, -5.0981, -5.5474, -5.6498, -4.2508, -3.6384, -4.9230,-5.5150, -3.9948, -5.5786, -4.2315, -4.3435, -5.1986, -4.8257, -4.2392,-4.6649, -5.5955, -5.2190, -4.8896, -5.6084, -5.1639, -5.6644, -4.2302,-4.6478, -5.2653, -5.4640, -4.5797, -6.1926, -4.3489, -5.2554, -4.2098,-5.5954, -4.5482, -4.8853, -5.3986, -4.9047, -4.2353, -4.8603, -5.6714,-5.5819, -5.5323, -4.5244, -3.7890, -5.6494, -5.6447, -4.6866, -4.5489,-5.4129, -5.2079, -5.7100, -5.2953, -4.0829, -5.3427, -4.6995, -5.5088,-5.3798, -5.3735, -4.4773, -5.0795, -5.4732, -5.3964, -5.0798, -4.6644,-5.7256, -6.7874, -4.4255, -4.7243, -4.0838, -4.5000, -5.7769, -5.8769,-5.8356, -5.6986, -4.8260, -4.9483, -3.8793, -4.4843, -4.7780, -2.8808,-5.0341, -5.5861, -4.7410, -4.6428, -6.2251, -4.2188, -3.2221, -5.4640,-5.5350, -3.3220, -5.3559, -5.2414, -5.0133, -3.9686, -5.3160, -4.1124]],device='cuda:0', grad_fn=<AddmmBackward0>)
预测结果是:Bananaquit

classes

['Bananaquit', 'Black Skimmer', 'Black Throated Bushtiti', 'Cockatoo']